Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Moving Toward Equitable Transit-Oriented Developments by Integrating Transit and Housing

193 views

Published on

Hongwei Dong, Cal State, Fresno

Published in: Education
  • Be the first to comment

  • Be the first to like this

Moving Toward Equitable Transit-Oriented Developments by Integrating Transit and Housing

  1. 1. Moving Toward Equitable Transit-Oriented Development by Integrating Transit and Housing Hongwei Dong, Ph.D. Department of Geography and City & Regional Planning California State University, Fresno hdong@csufresno.edu Friday Transportation Seminar Transportation Research and Education Center (TREC) @ Portland State University April 10, 2020 (Webinar) This study is preliminary and has not been peer-reviewed. Please do NOT cite.
  2. 2. About Myself • Associate Professor • Department of Geography and Planning at California State University Fresno (2011- present) • Research interests: housing and real estate, transportation and land use, & healthy cities • Ph.D. in Urban Studies from the Toulan School at PSU (2005-10) • Selected my research area after attending a Friday Transportation Seminar talk by Brian Gregor (ODOT research staff) in 2008 • Dissertation: developed a housing supply model to test Portland’s smart growth policies (Part of a larger IntegratedTransportation and Land Use Forecast Model)
  3. 3. About Myself q Maintained my interest in Portland’s urban growth and housing market while working in California • Dong, Hongwei. (forthcoming). Higher-density development for lower-cost housing? Understanding the multifamily housing market and the role of density in multifamily home prices. Journal of Planning Education and Research. • Dong, Hongwei, and J.Andrew Hansz. 2019. Zoning, density, and rising housing prices:A case study in Portland, Oregon. Urban Studies, 56(16): 3486-3503 • Dong, Hongwei. 2017.Transit induced neighborhood change and the affordability paradox of TOD. Journal ofTransport Geography, 63, 1-10. • Dong, Hongwei. 2016. If you build rail transit in suburbs, will development come? Journal of the American Planning Association, 82(4): 316-326.
  4. 4. My research on Portland’s TOD and Gentrification q Dong, Hongwei. 2017.Transit induced neighborhood change and the affordability paradox of TOD. Journal ofTransport Geography, 63, 1-10. Ø A longitudinal quasi-experimental design to examine five rail transit lines in suburban Portland and gentrification .
  5. 5. My research on Portland’s TOD and Gentrification q Major findings from this study Ø No consistent evidence for rail-transit-induced gentrification in suburban Portland. Ø No evidence that rail transit reduced home affordability Ø More changes in the neighborhoods served by the Eastside line (the oldest) § Attracted older and less-educated population § Experienced densification and faster increases of the share of rental units Ø Rail transit was more likely to be installed along low-income suburban neighborhoods
  6. 6. Outline 1. Background & Literature 2. Research Question & Study Area 3. Data, Measurement, & Method 4. Analysis 5. Finding 6. Conclusion
  7. 7. Background:TOD q Transit-oriented development (TOD) Ø Centered on transit (mainly rail transit) Ø Walkable and compact neighborhoods § Higher density § Mixed land use § Walkability Transit-oriented development (TOD) Source:The Next American Metropolis: Ecology, Community, and the American Dream by Peter Calthorpe (1993)
  8. 8. Background: Connecting Transit with Housing q Alleviate California’s housing affordability crisis via TOD? q Proposed SB50: upzoning near transit and jobs Ø Cities required to allow apartment buildings: § within a 1/2-mile of a rail transit station; § within a 1/4-mile of a high-frequency bus stop; or § within a “job-rich” neighborhood. Ø Upzone to allow buildings to be 45/55 feet tall Ø Reduce parking requirement significantly Source: https://medium.com/@Scott_Wiener/senator-wiener-introduces-zoning- reform-bill-to-allows-more-housing-near-public-transportation-and-3fb77b794004
  9. 9. Background: Connecting Transit with Housing q SB50 has been very controversial Ø Wealthy home owners: NIYMBY Ø Low-income tenants: gentrification Source: https://www.citylab.com/equity/2020/01/california-sb50-vote-affordable-housing-zoning-law- transit/605767/ Source: https://www.sandiegouniontribune.com/news/environment/sd-me-transit- gentrification-20180517-story.html
  10. 10. Literature: Connecting Transit with Housing q Property-value effects of rail transit Ø A well studied topic: housing transaction data are readily available Ø Majority of studies found significant and positive impacts § Benefit property owners § Justify the high cost of rail transit • Greater property tax base • More tax revenues for local governments
  11. 11. Literature: Connecting Transit with Housing q How about renters? Do they benefit from TOD? Ø A understudied topic: rent data are harder to obtain Ø Equity implications: § TOD premium is a burden instead of a benefit § Renters have lower income and more housing-burdened § Gentrification and displacement
  12. 12. Research Question q Today’s presentation focuses on the impacts of TOD on rents q Question 1: How much more do Californian renters have to pay to live in TODs? q Question 2: Does TOD rent premium vary: Ø Renters in different metro areas in California Ø Different dwelling sizes (studio, 1 bedroom, 2 bedroom, 3+ bedroom) Ø DifferentTOD types (urban TOD, suburbanTOD, &TAD)
  13. 13. Study Area q Rail transit stations in eight Californian metropolitan areas Ø 708 rail transit stations § Removed two funicular stations and 12 airport rail link stations Ø Use 694 rail transit stations for this analysis § San Francisco: 281 § Los Angles: 148 § San Diego: 83 § San Jose: 84 § Sacramento: 54 § Riverside (Inland Empire): 33 § Santa Rosa (Sonoma County): 6 § Oxnard (Ventura County): 5
  14. 14. Data & Measurement q Data sources Ø Rent: Craigslist.com Ø Rail transit:Transit Explorer 2 (thetransportpolitic.com) Ø Neighborhood social & built environments: § American Community Survey (ACS) 2014-18 § Longitudinal Employer-Household Dynamics (LEHD 2017) Ø Boundary and road: Census TIGER
  15. 15. Data & Measurement q Craigslist rent data: Ø Scraped 12/27/2019 – 01/31/2020 Ø Non-traditional data: may not be representative § Over-represent whiter, wealthier, and better-educated communities/groups Ø Advantages of Craigslist data § Crowd-sourced: comprehensive, large & free § More current than traditional data set (ACS & AHS) § Available at fine spatial scale (point level) § Richer information about the spot market Boeing, G., & Waddell, P. (2017). New insights into rental housing markets across the United States:Web scraping and analyzing Craigslist rental listings. Journal of Planning Education and Research, 37(4): 457-476. Boeing, G. (2019). Online rental housing market representation and the digital reproduction of urban inequality. EPA: Economy and Space, advanced online publication. Boeing, G, Wegmann, G., & Jiao J. (2020) Rental housing spot markets: how online information exchanges can supplement transacted- rents data. Journal of Planning Education and Research, advanced online publication.
  16. 16. Data & Measurement q Craigslist rent data are messy Ø Many duplicates § Landlords re-post/update their listings every few days to maintain visibility Ø Inaccurate/incomplete addresses information q Data cleaning is very time-consuming and tedious Ø 80% time on data cleaning & 20% time on data analysis q This analysis Ø 370,013 listings scraped Ø 73,775 used for analysis
  17. 17. Data & Measurement q Median rent: Craigslist over-represents higher-end rental units Metro Craigslist Jan. 2020 ACS 2014-18 AHS 2017 San Francisco $2,850 $1,687 $1,900 Los Angeles $2,160 $1,363 $1,400 San Diego $2,020 $1,465 n.a. San Jose $2,850 $1,996 $2,200 Sacramento $1,599 $1,084 n.a. Riverside $1,652 $1,119 $1,100 Oxnard $2,150 $1,595 n.a. Santa Rosa $2,227 $1,412 n.a.
  18. 18. Data & Measurement q Dwelling size: Craigslist over-represents units with one-bedroom but under-represents units with 3+ bedrooms Metro Size Craigslist (%) ACS 2014-18 (%) AHS 2017 (%) San Francisco studio 10.0 11.7 7.2 1 bedroom 39.1 30.7 34.8 2 bedrooms 35.3 34.7 35.6 3+ bedrooms 15.6 22.9 22.4 Los Angeles studio 10.6 10.2 5.5 1 bedroom 39.4 31.1 35.8 2 bedrooms 37.8 38.3 39.2 3+ bedrooms 12.2 20.5 19.5 San Jose studio 6.0 7.8 3.0 1 bedroom 37.8 27.7 31.0 2 bedrooms 39.6 37.5 37.5 3+ bedrooms 16.6 27.1 28.5 Sacramento studio 3.4 4.2 1.1 1 bedroom 29.0 22.8 16.9 2 bedrooms 40.8 38.0 40.9 3+ bedrooms 26.8 35.0 41.1 San Diego studio 6.3 6.1 1 bedroom 32.5 25.1 2 bedrooms 43.4 42.0 3+ bedrooms 17.8 26.8 N/A
  19. 19. Data & Measurement q Neighborhood: Craigslist listings tend to be in neighborhoods with more jobs, newer homes,Whites & Asians, and higher income. 0.00 0.50 1.00 1.50 2.00 2.50 Share of Hispanic pop. Share of Black pop. Share of SFH Pop. density Housing built before 1940 Share of Asian pop. Meidan rent Meidan Household income Share of White pop. Housing built after 1999 Job density Mixed use Ratio: Craiglist/Regional average
  20. 20. Data & Measurement q Measuring TOD: within 0.5-mile street network distance Ø TOD (treated) units: rental units within 0-0.5 mile (yellow area) Ø Non-TOD (control) units: rental units > 1.0 mile (outside of blue area)
  21. 21. Method:Why Use PSM? q Three potential approaches to tease out the effect of TOD on rents Ø OLS hedonic model does not address two critical issues § Spatial autocorrelation § Self-selection Ø Spatial regression model (spatial lag, spatial error, & Durbin/mixed) § Assumes the spatial relationship is known § Does not address the self-selection problem Ø Propensity score matching (PSM)
  22. 22. Method:Why Use PSM? q I adopt the propensity score matching (PSM) method Rental units in TODs (within 0.5-mile network distance) Rental units in transition area (within 0.5-1.0 mile network distance) Rental units not in TODs (> 1.0-mile network distance) PSM Rental units in TODs (treatment) Rental units not in TODs (control) Similar housing attributes Similar neighborhood environment Similar location Price difference (average treatment effect)
  23. 23. Method:Why Use PSM? q Why use propensity score matching (PSM)? Ø Address the self-selection bias by identifying a control group Ø Spatial autocorrelation is not a concern Ø Study design is determined before analysis, like randomized controlled trial
  24. 24. Method: PSM q PSM: allow replacement or not? Ø Without replacement: § One untreated case can be used only once as a control case § Higher-quality matching but some treated cases may not be matched Ø With replacement: § One untreated case can be used as a control for multiple treated cases § Lower-quality matching but almost all treated cases can be matched
  25. 25. Analysis: Identify Control Cases q In the following, I will report the results from PSM without replacement because of their higher quality of matching q Not every TOD rental unit can be matched with a control unit when replacement is not allowed Treated Control Treated Control All 12,863 28,134 12,863 28,134 Matched 7,446 7,446 12,863 6,097 Unmatched 5,417 30,688 0 32,037 Matched without replacement Matched with replacement Treated = TOD rental units; control = non-TOD rental units
  26. 26. Method: PSM q Control variables (covariates in PSM) Ø Housing attributes (bedroom, bathroom, building structure) Ø Neighborhood environment § Land use: activity density (pop. & job) & mixed land-use § Housing stock: shares of rental units, share of single-family homes, & age (new homes built since 2000, & old homes built before 1940) § Social environment: median household income & shares of Black and Hispanic pop. Ø Location: distance to CBD & specific metropolitan area
  27. 27. Method: PSM q Balance diagnostics: how similar are treatment and control groups? Ø Paired t-tests: compare their mean values § Widely used in literature § Problem: sensitive to sample size (larger sample size becomes a “disadvantage”?) Ø Standardized difference (Austin 2011) § Not sensitive to sample size § Standardized difference <0.1 indicates negligible difference Ø Variance ratio: compare distribution (Austin 2011) Austin, P.C. (2011).An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behavioral Research, 46: 399-424.
  28. 28. Analysis: Identify Control Cases q The covariates of the treatment and control groups (7,446 pairs ) are well balanced Mean value: TOD units Mean value: non-TOD units Mean difference Standardized mean difference Bedroom 1.50 1.51 -0.01 -0.011 Bathroom 1.36 1.36 -0.01 -0.019 Neighborhood environment: Activity density (pops & jobs per acre) 41.22 38.50 2.72 0.002 Mixed land use 0.33 0.32 0.01 0.022 Median household income ($1000) 77.50 77.86 -0.36 0.000 Share of Black pop. (%) 6.46 6.32 0.15 0.002 Share of Hispanic pop. (%) 27.64 27.41 0.23 0.000 Neighborhood housing stock: Share of SFHs (%) 26.52 28.31 -1.79 -0.003 Share of rental housing (%) 74.55 72.97 1.58 0.004 Share of housing built since 2000 (%) 20.86 22.06 -1.20 -0.002 Share of housing built before 1940 (%) 16.59 14.84 1.75 0.004 Location: distance to CBD 10.96 11.71 -0.74 -0.008 In Los Angeles (yes=1) 0.33 0.34 0.00 -0.009 In San Francisco (yes=1) 0.25 0.23 0.02 0.048 In San Jose (yes=1) 0.11 0.10 0.01 0.042 In San Diego (yes=1) 0.17 0.20 -0.02 -0.061 In Sacramento (yes=1) 0.08 0.07 0.01 0.022
  29. 29. Analysis: Identify Control Cases q What types ofTOD units were not matched when replacement is not allowed? Ø In neighborhoods with higher levels of density and mixed use Ø In neighborhoods with newer and older housing Ø Closer to CBD Ø In San Francisco
  30. 30. Finding:TOD Rent Premium in CA q TOD premium in California: $127 TOD units (treated) Non-TOD units (control) N 7,446 7,446 Mean monthly rent $2,545 $2,418 TOD premium **statistically significant at the 1% level Matched without replacement $127** (5.3%)
  31. 31. Finding:TOD Rent Premium by Dwelling Size q TOD premiums are higher for rental units of larger sizes Ø Studio 4.6%; 1-bedroom: 4.0%; 2-bedroom: 6.8%; 3-bedroom: 7.6% $85 $86 $180 $242 STUDIO 1-BEDROOM 2-BEDROOM 3+ BEDROOM TOD premium (Estimated by PSM without replacement)
  32. 32. Finding:TOD Rent Premium by TOD Types q Group 694 rail transit stations into 3 clusters (via cluster analysis) Ø UrbanTOD, Ø SuburbanTOD, & Ø TAD (transit-adjacent development) Urban TOD (N=134) Suburban TOD (N=339) TAD (N=221) Distance to CBD (mile) 1.5 6.4 16.1 Population density (persons/acre) 54.9 19.8 7.1 Job density (jobs/acre) 120.5 15.7 8.7 Street density (mile/acre) 0.038 0.029 0.019 Service area (acre) 432.5 370.3 228.7 Metro areas: Bay Area 79.9% 58.1% 30.3% Los Angeles 7.5% 29.2% 34.8% Sacramento 9.7% 4.1% 12.2% San Diego 3.0% 8.6% 22.6% Total 100.0% 100.0% 100.0%
  33. 33. Finding:TOD Rent Premium by TOD Types q TOD premiums for three types of TODs Ø SuburbanTOD > UrbanTOD Ø EvenTAD generates some premium $128 $153 $76 URBAN TOD SUBURBAN TOD TAD TOD premium (Estimated by PSM without replacement)
  34. 34. Finding:TOD Rent Premium in Metro Areas q TOD rent premium in three major regions Ø LA: none; Bay area: 7.0%; & San Diego: 8.2%. -$23 $202 $169 GREATER LA BAY AREA SAN DIEGO TOD premium (Estimated by PSM without replacement) *TOD premium in the greater LA region is not statistically different from zero
  35. 35. Caveat q Cross-sectional data q Craigslist data over-represent higher-end rental units in well-off neighborhoods q Hard to find matches/control cases for TOD rental units that are: Ø Small-sized (studios & 1-bedroom) Ø Located in central-city neighborhoods, particularly those in San Francisco q TOD premium is less certain for central-urban rental units Ø This is a hidden (but important) issue when running a hedonic regression
  36. 36. Conclusion q Craigslist data are very useful Ø Until local governments systematically collect rent data and make them public (like what they do with housing transaction data) Ø Portland to require landlords to register rental properties to the city by 2020 q PSM shows advantages over and more transparency than the hedonic regression method
  37. 37. Conclusion q The average TOD rent premium is $127 Ø About 5.3% of average rent in TODs q TOD rent premium is greater for larger rental untis in absolute value and in percentage q TOD rent premium varies in different metro areas Ø None in LA and around 7.0% in the Bay area, and 8.2% in San Diego q Suburban TOD rent premium is greater than urban TOD rent premium
  38. 38. Conclusion q TOD & gentrification Ø The threat of gentrification is real, at least in the Bay area and San Diego Ø TOD rent premium could worsen the housing affordability crisis Ø However, the overall effect depends on how much renters could save on transportation expenditures q Equity implications Ø Renters vs. homeowners § Renters are in a more disadvantaged position, compared to homeowners Ø A windfall for landlords/housing investors § They may have to pay higher property taxes. Is this enough? § We may need better value-capture mechanisms
  39. 39. Conclusion q Next step: Ø Estimate the transportation cost saving effects ofTOD Ø Compare TOD rent premium with transportation cost savings Living in TODs Higher rent Drive less Own fewer cars Use transit more Overall H+T expenditures? Fewer expenditures on cars More expenditures on transit Overall transportation expenditures* *Travel time cost is not considered.

×